A NEW WAY TO LOOK AT ALUMNI WEB SURVEY DATA
By Peter B. Wylie, with John Sammis
I have known Ray Satterthwaite since 1999. At the time he was working at McGill
University in Montreal. He had called me about an article I'd written for CASE
CURRENTS in May of that year. It was about data mining. We've been friends and
colleagues ever since.
Anyway, Ray has a web survey company that does a lot of work for secondary
schools and universities. He and I have talked for years about some of the
methodological problems with all web surveys. Ray is keenly interested in trying
to find ways around these problems, especially non-response. Right now John
Sammis and I are working with him on a technique to estimate how nonresponders might have filled out a survey. We're excited about the technique but
feel it's too soon to roll it out for general scrutiny. When we do, it'll be
controversial. That's okay; controversy can lead to new and better ideas.
In this piece we want to use some of Ray's data from a recent web survey to show
several things you can do with one of your own web surveys if ... if you're willing
to put some gentle pressure on both your IT folks and the vendor who performs
your survey(s).
Why go to the extra trouble of doing what we're suggesting? Simply put, we think
it's smart to match up survey data with at least some of the plentiful data you have
on alums who respond to your survey as well as those who don't. Foraging through
these data can lead to useful insights. None of us in advancement can do too much
of this kind of analysis.
Here's a summary of the steps taken to make it possible to write this piece:
1. All alumni who could be reached by email were invited to complete the
survey by going to a secure website. The total number of alumni considered
for the survey was 204,435.
2. Several reminders were sent out over the course of several weeks to alums
who had not completed the survey.
3. Once the survey was closed, 6,281 usable survey forms were stored so that
each respondent could be matched by a unique ID number to other data
stored by the school for that ID number.
1 4. A file was provided (via Ray) to John Sammis who matched up an overall
measure of alumni engagement derived from the survey with a long list of
fields such as:
•
•
•
•
•
•
•
•
Total lifetime giving
Giving over the last five fiscal years
Current member of the alumni association (yes/no)
Marital status
Mobile phone listed (yes/no)
Zip code
Number of degrees
Undergrad major
5. Sammis/Wylie began analyzing some of these data.
Let’s jump right in. Take a peek at Table 1. At first glance, it looks kind of busy.
But once you’ve had a chance to study it, we think some interesting stuff will
emerge.
Table 1: Frequency Breakdown of 204,435 Alumni by Class Year and Level of
Engagement
DECILE
GROUPS
BEFORE
1981
Non Responders 1 2 3 4 5 6 7 19811993
19942005
20062014
52,360 166 48,532 131 52,100 139 51,143 193 155 144 146 183 132 144 152 199 148 144 146 191 127 132 152 217 123 112 172 221 113 129 142 244 2 8 9 10 105 130 120 273 106 119 135 268 96 90 119 323 For example, look below at this replica of Table 1. Focus on the numbers in the
green font. They show the number of non-responders for each of the class year
quartiles. What we’re seeing here is pretty clear. Most of the 204,435 alums who
might have responded to the survey did not respond. Regardless of how long
they’ve been out of school.
DECILE
GROUPS
BEFORE
1981
Non Responders 1 2 3 4 5 6 7 8 9 10 19811993
19942005
20062014
51,089 166 47,257 131 50,677 139 49,131 193 155 144 146 183 132 144 152 199 148 144 146 191 127 132 152 217 123 112 172 221 113 129 142 244 105 130 120 273 106 119 135 268 96 90 119 323 3 Now let’s look at the responders. In particular, let’s focus on the responders from
the oldest class year quartile and the youngest class year quartile. These alums are
now in the green font in another replica of Table 1. DECILE
GROUPS
BEFORE
1981
Non Responders 1 2 3 4 5 6 7 8 9 10 19811993
19942005
20062014
51,089 166 47,257 131 50,677 139 49,131 193 155 144 146 183 132 144 152 199 148 144 146 191 127 132 152 217 123 112 172 221 113 129 142 244 105 130 120 273 106 119 135 268 96 90 119 323 There’s something interesting going on here, but it’s a little tough to see it in the
table. It’s easier to see in Figure 1. Notice that the number of young alums
increases as the engagement deciles increase. Just the opposite occurs with the
oldest alums. That is, the younger alums are saying they are more engaged than the
older alums.
4 Fig. 1: Frequency of Alums by Engagement Level for the Oldest and Youngest Class Year QuarNles 350 300 250 200 150 100 50 0 1 2 3 4 5 6 7 8 9 10 Engagement Deciles BEFORE 1981 2006-‐2014 Let’s shift gears. Let’s look at a different kind of problem that Ray's data has
allowed us to bite into: The relationship between what alums say about their alma
maters and what they give to their alma maters.
"Ah ... hasn't somebody already looked at that?" you might be asking.
Well, sort of. The trouble is that studies reporting the relationship between alumni
giving and attitudes are almost all based on survey data. That is, the giving
variables in these studies are an alum's answers to questions about what, how
much, how often, etc. the person has given to the school.
Call us crazy, but we’re not real comfortable with basing someone's giving on what
they say they've given. It's sort of like the weight people put on their driver's
licenses. No offense, but we'll stick with what the scale says.
Our clowning around aside, check out Kevin MacDonell's piece from November
2010: “Survey says … beware, beware!”1 Kevin says, "Frankly, if I asked survey
takers how many children they have, I wouldn't trust the answers."
1
https://cooldata.wordpress.com/2010/11/18/survey-‐says-‐beware-‐beware/ 5 In the rest of the piece we’ll lay out a series of tables and graphs that show the
relationship between the engagement deciles and:
•
•
•
•
Recency of giving
Frequency of giving
Magnitude of giving
Current membership in the school’s alumni association
You’ll see a lot of detail in these tables and charts. Take as much or as little time
looking through them as you’d like. All along the way we’ll provide our sense of
what we see going on in this material.
If all the detail seems a bit much, you might want to step back and just ask yourself
a general question: Is this the kind of analysis we should be doing with our own
survey data, regardless of whether we do it ourselves or have our vendors do it
for us?
Recency of Giving
There are lots of ways to operationally define recency of giving. Here we’ve
defined it as whether or not an alum has given anything at all in the last five fiscal
years.
Because Table 2 is the first of a series of tables with similar formats, we’ll go
through it in more detail than the rest of the tables to make sure we’re being clear.
Notice that all the cell entries in the table are percentages. For example, if you look
in the cell for non-responders who graduated before 1981, you’ll see that 17.5% of
them gave anything at all in the last five fiscal years. Alums who landed in the
third decile and graduated between 2006 and 2014? 12.6% of them gave anything
at all during this same period. One more: Alums who landed in the tenth decile and
graduated before 1981? 71.9% of them gave anything at all during this same
period.
Notice that Figure 2 shows the same information as you see in the table, except the
non-responders are not shown. (We left them out because we decided including
them would make the chart too busy.)
6 Table 2: Percentage of Alums Who Have Given Anything at All in the Last
Five Fiscal Years by Class Year Quartile and by Engagement Level
DECILE
GROUPS
Non Responders 1 2 3 4 5 6 7 8 9 10 BEFORE
1981199420061981
1993
2005
2014
17.5% 14.6% 9.9% 8.1% 22.3% 13.0% 5.8% 7.3% 33.5% 27.8% 17.8% 13.7% 36.4% 24.3% 21.7% 12.6% 43.2% 34.7% 18.5% 11.5% 52.0% 40.9% 24.3% 12.0% 49.6% 43.8% 29.1% 14.9% 51.3% 47.3% 30.3% 17.6% 56.2% 43.8% 39.2% 20.1% 73.6% 63.0% 42.2% 21.6% 71.9% 62.2% 41.2% 19.5% 7 Fig. 2: Percentage of Alums Who Have Given Anything at All in the Last Five Fiscal Years by Class Year QuarNle and by Engagement Level 80.0% 70.0% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% 1 2 3 BEFORE 1981 4 5 1981-‐1993 6 7 1994-‐2005 8 9 10 2006-‐2014 A couple things stand out for us in these data:
Survey responders are better recent givers than are non-responders. We’re not
surprised by this. Whenever we’ve been able to look at the giving rates of alums
who’ve responded to a survey versus those who haven’t, the responders (on
average) come out ahead. This phenomenon is particularly pronounced for alums
who graduated before 1981. In this group even the alums in the lowest decile
(people whose survey responses say they’re not engaged) have a higher rate of
recent giving than non-responders: 22.3% versus 17.5%.
There is clearly a positive relationship between engagement and recency of
giving, except for the young alums. It’s easier to see this relationship in Figure 2
than in Table 2. The blue and red and gray lines (representing the oldest three
quartiles of alums) go up pretty smoothly. This is far less the case for the youngest
alums represented by the yellow line. In plain English we might say: What the
youngest alums tell us in the survey about their engagement corresponds far less
closely to their recent giving than it does for their older counterparts.
8 Frequency of Giving
We’ve defined frequency of giving as the number of gifts an alum has made over
the last five fiscal years.
The cell entries in Table 3 are not percentages like we saw in Table 2. They are
means (arithmetic averages). So if you look in the cell for non-responders who
graduated before 1981, you’ll see that the mean number of gifts over the last five
years for an alum is .40. That might be a little confusing. How can someone make
four tenths of a gift? Maybe an easier way of looking at these numbers is to think
of them in terms of 100 alums. Four tenths of a gift for one alum is the same as 40
gifts for 100 alums. Now look at the cell entry for alums who graduated before
1981 for the tenth engagement decile. It’s 2.39. So applying the same logic, this is
the same as 239 gifts for 100 alums. A big difference.
Notice that Figure 3 shows the same information as you see in the table. Again, the
non-responders are not shown.
Table 3: Mean (Average) Number of Years of Alumni Giving Over the Last
Five Fiscal Years by Class Year Quartile and by Engagement Level
DECILE
BEFORE
198119942006GROUPS
1981
1993
2005
2014
Non Responders 0.40 0.30 0.18 0.12
0.44 0.24 0.10 0.09 1 0.83 0.58 0.34 0.18 2 0.86 0.53 0.55 0.14 3 1.01 0.80 0.45 0.16 4 1.39 0.89 0.57 0.16 5 1.45 1.13 0.60 0.24 6 1.49 1.12 0.64 0.29 7 1.56 1.06 0.92 0.33 8 2.09 1.73 1.09 0.32 9 2.39 2.02 1.15 0.31 10 9 Fig.3: Mean (Average) Number of Years of Alumni Giving Over the Last Five Fiscal Years by Class Year QuarNle and by Engagement Level 3 2.5 2 1.5 1 0.5 0 1 2 3 BEFORE 1981 4 5 1981-‐1993 6 7 1994-‐2005 8 9 10 2006-‐2014 Since the patterns for frequency of giving are quite similar to what we saw with
recency of giving, let’s move on to magnitude of giving.
10 Magnitude of Giving
For magnitude of giving we chose to compute the median lifetime giving for each
cell in Table 4.
We think the numbers in this table and in the corresponding chart are impressive.
More about that in a moment. Now let’s just be sure we’re clear on what the
numbers mean. For example:
• The median for non-responding alums who graduated before 1981 is $5.
That means that half of these alums have given five dollars or less lifetime.
The other half have given five dollars or more lifetime.
• The median for alums in the tenth engagement decile who graduated before
1981 is $1,818. That means that half of these alums have given $1,818 or
less lifetime. The other half have given $1,818 or more lifetime.
• Any cell that contains a median of $0 means that over half of the alums in
that cell have given nothing at all lifetime.
Table 4: Median Alumni Lifetime Giving by Class Year Quartile and by
Engagement Level
DECILE
GROUPS
BEFORE
1981
19811993
19942005
Non Responders $5 $0 $0
$23 $13 $0 1 $93 $34 $0 2 $153 $25 $0 3 $195 $73 $0 4 $190 $100 $0 5 $250 $100 $23 6 $600 $150 $17 7 $388 $178 $30 8 $906 $325 $35 9 $1,818 $518 $35 10 20062014
$0 $0 $0 $0 $0 $0 $0 $0 $0 $0 $0 11 Fig. 4: Median Alumni LifeNme Giving by Class Year QuarNle and by Engagement Level $2,000 $1,800 $1,600 $1,400 $1,200 $1,000 $800 $600 $400 $200 $0 1 2 3 BEFORE 1981 4 5 1981-‐1993 6 7 1994-‐2005 8 9 10 2006-‐2014 As we said earlier, we’re impressed with what we see in Table 4 and Figure 4.
How come? Look at those older alums, the ones who graduated before 1981. Focus
on the ones who didn't respond? The ones whose median lifetime giving is five
bucks? Well, that’s just … choose any pejorative adjective you like. Now look at
how the median steadily rises (with one exception) as expressed engagement
increases. Finally, look at the alums in that group who say they are most engaged.
Those guys and gals have really put their money where their mouths are. Wow!
12 Current Membership in the School’s Alumni Association
The numbers in the cells of Table 5 indicate the percentage of alums who are
currently dues paying members of the school’s alumni association. (We won’t bore
you with, once again, explaining how to interpret the table and the graph.)
Table 5: Percentage of Alums Who Are Current Members of the Alumni
Association by Class Year Quartile and by Engagement Level
DECILE
GROUPS
Non Responders 1 2 3 4 5 6 7 8 9 10 BEFORE
19811981
1993
29.4% 20.3% 199420062005
2014
10.4% 7.3% 10.2% 5.3% 1.4% 3.1% 10.3% 11.1% 5.5% 1.6% 19.7% 6.9% 6.6% 1.5% 23.0% 16.7% 7.5% 5.8% 24.4% 15.2% 11.8% 4.6% 27.6% 24.1% 11.6% 3.2% 38.9% 26.4% 12.7% 7.8% 42.9% 22.3% 24.2% 7.0% 47.2% 38.7% 25.9% 10.4% 64.6% 51.1% 34.5% 13.3% 13 Fig. 5: Percentage of Alums Who Are Current Members of the Alumni AssociaNon by Class Year QuarNle and by Engagement Level 70.00% 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% 1 2 3 BEFORE 1981 4 5 1981-‐1993 6 7 1994-‐2005 8 9 10 2006-‐2014 Much of what we see in Table 5 and Figure 5 is similar to the patterns we saw
when we looked at the three different types of giving (recency, frequency, and
magnitude). However, a couple of things caught our eye:
Except for the youngest class year quartile, there is quite a pronounced difference
between alums in the top engagement decile and everybody else. This is
particularly true for the oldest class year quartile. Look at the percentage of alums
who are current members of the association in the tenth decile: 64.6%. That’s way
more than the percentage for any other cell in the table. These are the same folks
whose lifetime giving dwarfed everybody else’s. These people say they’re very
engaged; what they give in raw dollars and pay in dues makes us really want to
believe them.
14 Overall Comments
What we’ve done here is show you a smidgeon of the analysis that could have been
done with the survey data from one school that has been matched with data from
that school’s alumni database. We hope you’ve found this interesting.
But the big reason we showed you this analysis is because of a strongly held belief.
A belief that neither alumni survey vendors nor their higher education clients are
doing the kind of data matching that makes this kind of analysis possible.
And what are some of the consequences of this lack of analysis? Here are just a
couple that come to mind:
• Vendors can convince clients that a relatively small group of responders is
not much different from a very large group of non-responders, when that is
patently not the case. These two groups can differ hugely with respect to a
host of important variables like giving, time out of school, gender, living
location, association membership and on and on the list can go.
• Clients don’t have the opportunity to hone in on certain groups of responders
who express a high degree of engagement in a survey and back that
expression up with strong giving and volunteer involvement. Take the group
of responding alums who graduated before 1981 and whose median lifetime
giving is over $1,800. There are only 90 of these folks. Wouldn’t it be great
to ask each of these alums for permission to interview them in person or
over the phone on why they are so engaged with the school? Many of them
would decline the chance to be interviewed. But many would agree. Just
imagine the stories they could tell. We’d be hanging on every word.
To be continued …
15